Brain and Language 85 (2003) 245–261
www.elsevier.com/locate/b&l
PET activation studies comparing two speech tasks widely used
in surgical mapping
Diana Vanlancker-Sidtis,a,* Anthony Randal McIntosh,b and Scott Graftonc
b
a
Department of Neurology, University of Southern California, Los Angeles, CA, USA
The Rotman Research Institute of Baycrest Centre, University of Toronto, Toronto, Ont., Canada
c
The Center for Cognitive Neuroscience, Dartmouth College, Hanover, NH, USA
Accepted 29 October 2002
Abstract
‘‘Automatic’’ speech, especially counting, is frequently preserved in aphasia, even when word production is severely impaired.
Although brain sites and processes for automatic speech are not well understood, counting is frequently used to elicit fluent speech
during preoperative and intraoperative cortical mapping for language. Obtaining both behavioral and functional brain imaging
measures, this study compared counting with a word production task (generation of animal names), including non-verbal vocalizations and quiet rest as control states, in normal and aphasic subjects. Behavioral data indicated that normal and aphasic groups
did not differ in counting or non-verbal vocalizations, but did differ significantly in word production (‘‘naming’’ animals). Functional brain imaging results on normal subjects using partial least squares analysis of PET rCBF images revealed three significant
latent variables (LVs): one for naming and vocalizing, identifying bilateral anterior areas, with left predominating over right; a
second LV for naming, identifying left and right frontal and temporal areas. For the third, only marginally significant LV, which
was associated with automatic speech alone (counting), right and subcortical sites predominated. For patients, two LVs emerged,
identified with naming and vocalization, and corresponding to a variety of cerebral sites; the analysis failed to find a specific latent
variable for counting. A comparison between group data for normal subjects and patients suggested that the naming, counting, and
vocalization tasks were performed differently by the two groups. These results suggest that word generation as a verbal task is more
likely to elicit activity in classical language areas than counting. Further studies are suggested to better understand differences
between neurological substrates for non-propositional and automatic speech.
Ó 2003 Elsevier Science (USA). All rights reserved.
Keywords: Speech; Functional mapping; Language; Human; Cerebral blood flow; Automatic speech
1. Introduction
Within the adult human brain, the left cerebral
hemisphere plays a crucial role in most aspects of oral
and written language. From clinical–pathological studies in brain-injured patients, it is well known that
damage to different regions of the left hemisphere causes
aphasic syndromes with different patterns of language
impairments. Extensive damage can cause global aphasia, in which most aspects of speech production and
*
Corresponding author. Present address: Department of Speech
Pathology, NYU School of Education, 719 Broadway, Suite 200 New
York, NY 10003, USA. Fax: 1-212-995-4356.
E-mail address: [email protected] (D. Vanlancker-Sidtis).
comprehension are severely impaired. Many such patients, however, retain striking areas of speech competency. Although unable to generate novel words or
sentences, these patients produce serial speech such as
counting, other lists (e.g., days of the week), and expletives (Van Lancker & Cummings, 1999), recite of
pledges, nursery rhymes and other well-known texts,
and sing familiar songs, with ease, fluency, and normal
articulation (Van Lancker, 1988, 1993). Similarly, patients with BrocaÕs aphasia, whose sparse, effortful
speech output is usually limited to single, poorly articulated nouns and verbs, can, under limited circumstances, produce normal sounding ‘‘subsets’’ of speech
often referred to as ‘‘automatic’’ speech (Albert & HelmEstabrooks, 1988; Espir & Rose, 1970). As Benson
0093-934X/03/$ - see front matter Ó 2003 Elsevier Science (USA). All rights reserved.
doi:10.1016/S0093-934X(02)00596-5
246
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
(1979) states, ‘‘The [Broca] patient often performs considerably better in automatic activities such as counting,
or naming the days of the week or the months; even the
articulation often improves greatly with these automatic
verbal tasks. It is striking to note how well a patient
articulates when reciting a series but cannot articulate
the same words correctly in a repetition task’’ (p. 67).
The same discrepancy is observed in fluent and anomic
aphasia, whereby semantic and word finding deficits
contrast with normal production of counting and other
overlearned speech.
Although this dissociated speech ability has been
recognized for well over a century (Broca, 1865; Code,
1982, 1989; Critchley, 1962, 1970; Henderson, 1985,
1987; Jackson, 1878; Van Lancker, 1988, 1993), it remains uncertain whether residual, intact utterances depend on neural structures within undamaged regions of
the left cerebral hemisphere, on cortical structures in the
nominally non-dominant right hemisphere, or on subcortical nuclear areas such as the basal ganglia (or on
some combination of these). Because the types of
‘‘automatic’’ speech are varied and they often appear to
be used ‘‘voluntarily,’’ this category is better referred to
as ‘‘non-propositional’’ speech. It is not known whether
these two apparently disparate speech functions—nonpropositional and propositional speech—are subserved
by common or different underlying structures in the
normal brain.
Lum and Ellis (1994) studied ‘‘non-propositional’’
speech tasks such as counting 1–10, reciting the days of
the week, months of the year, nursery rhymes, and repeating familiar phrases in patients with aphasia.
Overall performance differences were found for counting, and six of 28 patients tested showed a clear
non-propositional advantage. A similar investigation
revealed superior sentence completion for familiar idiomatic expressions compared with propositional expressions (Van Lancker & Bella, 1996). In a survey of
recurrent utterances preserved in severely aphasic persons, Code (1982) reported numbers, expletives, and
familiar expressions, categories pertinent to this study.
Similar results were reported for German-speaking
aphasic patients by Blanken (1991) and colleagues
(Blanken, Wallesch, & Papagno, 1990).
Although much has been written about preserved
‘‘automatic speech,’’ we know of only one study of its
impairment (Speedie, Wertman, TaÕir, & Heilman,
1993). Following a stroke to the right basal ganglia, a
75-year-old, right-handed man was unable to recite well
known prayers, count 1–20, or sing familiar songs.
Another study observed physical features in aphasic
speakers engaged in automatic and propositional tasks.
Measurement of mouth opening sizes comparing
production of automatic and propositional speech in
patients with left hemisphere stroke revealed greater leftsided mouth openings for recitation and singing, again
implicating the right hemisphere (Graves & Landis,
1985). These studies on preserved and impaired automatic speech in persons with brain damage implicate
right hemisphere cortical and subcortical structures in
production of non-propositional speech.
The question of the neurological structures underlying non-propositional versus propositional speech is of
importance and interest for at least three reasons. It is
important, first, for our understanding of basic brain
structures underlying normal speech ability. Secondly,
to develop viable theories of compensation and rehabilitation following language loss due to brain damage,
this information is important, because intact neurological structures subserving overlearned speech may have
different properties. For example, the brain structures
subserving residual speech may have constraints and
limitations that should be known by the rehabilitation
specialist designing a treatment plan. Third, this information will help lead to better informed choices of
presurgical and intrasurgical speech-mapping tasks to
delineate specific brain regions in the treatment of conditions such as epilepsy (see Lebrun & Leleux, 1993, for
review).
Regarding the third rationale for this study, selection
of cortical speech-mapping tasks, two different tasks are
commonly utilized during pre- and intraoperative cortical speech mapping when surgical excision may impinge on speech and language areas: (1) Counting
(Ojemann, 1983, 1994, 1995; Penfield & Roberts, 1959)
as a method for eliciting continuous, fluent motor
speech and (2) word retrieval using confrontation
naming (Berger & Ojemann, 1992; Mateer, 1983; Penfield & Roberts, 1959) as a vehicle for probing semantic
processes of word retrieval.
Some attempts to replace cortical mapping with
fMRI (Tomczak et al., 2000) and PET functional imaging studies (Hunter et al., 1999) have been undertaken. In the Tomczak et al. (2000) study, preoperative
location of motor areas using fMRI was more successful
than mapping of language areas, which is to be expected. However, reservations regarding the limitations
of both PET and fMRI as clinical tools in neurosurgery
for localization of cognitive function have been expressed (Fried, Nenov, Ojemann, & Woods, 1995, p.
860). The low statistical power of these methods raises
the possibility for false negative results, a situation that
is particularly problematic when the clinician is attempting to establish that a particular brain area does
not support a particular function such as memory or
language. Our goal in this paper is not to establish a
clinical test with imaging. It is more modest: to contribute to a better understanding of the neuroanatomical
substrates underlying these tasks in the normal, intact
subject and in the aphasic patient. Convergent information leading to knowledge about whether disparate
neurological structures subserve overlearned serial
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
speech versus propositional speech may aid in selecting
the optimal intraoperative speech-mapping procedures
for identifying speech/language cortex.
Studies using PET imaging in normal and neurologically impaired subjects have begun to explore speech/
language processing and verbal memory. Considerable
variability in findings has resulted from studies of language functions in normal subjects, with criticism of
method and technique appearing in the literature
(Demonet, Wise, & Frackowiak, 1993; Haxby, Grady,
Ungerleider, & Horwitz, 1991; Klein et al., 1997; Steinmetz & Seitz, 1991). Study design and method of data
acquisition are likely to influence findings in language
studies (Lange et al., 1999; Price et al., 1996). Complexity in this paradigm arises from individual variability
in subjects, significant variation in levels and structure of
language tasks, and questionable appropriateness of
analysis methods involving subtraction and other algorithms with loss of information through extensive averaging of data (Friston, Frith, Liddle, & Frackowiak,
1991). Questions arise about the lack of coherence of
PET findings with classic lesion models, in particular the
frequently reported involvement of the right hemisphere
in language functions traditionally believed to be
strongly lateralized to the left hemisphere (Sidtis, 2000;
Sidtis, Anderson, Strother, & Rottenberg, 1998). However, a number of consistent and reliable findings regarding brain mapping of speech functions have been
reported (e.g., Howard et al., 1992; Paus, Petrides, Evans, & Meyer, 1993; Petersen & Fiez, 1993; Petersen,
Fox, Posner, Mintun, & Raichle, 1988; Raichle, 1991;
Votaw et al., 1999) contributing to a foundation for
developing a viable model of language function in the
normal and diseased brain. A number of studies have
identified left temporal (Damasio, Grabowski, Tranel,
Hichwa, & Damasio, 1996) and/or left inferior frontal
lobe areas during semantic and phonological behavioral
tasks. Studies of continuous speech production using
PET have shown a significant relationship between
speaking rate and activity in the left inferior frontal area
and right basal ganglia (Sidtis et al., 2001), while fMRI
studies reported activation of inferior parietal as well as
temporal cortex (Kircher, Brammer, Williams, &
McGuire, 2000; see Wise, Hadar, Howard, & Patterson,
1991 and Poldrack et al., 1999 for overviews of language
studies; also Fiez et al., 1995; Vandenberge, Price, Wise,
Josephs, & Frackowiak, 1996; see Cabeza & Nyberg,
2000, for review of imaging studies of cognition).
Fewer studies have assessed language function in
brain damaged patients. Celesia et al. (1984) reported
larger areas of depressed rCBF than appeared anatomically on the CT scan; remote effects were also seen.
Metter et al. (1990) found hypometabolism in the left
superior temporal areas in nearly all aphasic subjects,
regardless of their diagnosis. Reading aloud and
speaking of words was used successfully to map lan-
247
guage areas in a single patient with an arteriovenous
malformation (AVM), whereby the baseline PET scan
first identified the location of the AVM (Leblanc,
Meyer, Bub, Zatorre, & Evans, 1992). A study of recovery of language function post-stroke indicated increased activation of newly recruited areas when
language areas remain non-functional (Karbe et al.,
1998).
Keeping in mind the limitations of the PET activation
paradigm, our protocol was designed to add to this
growing foundation of knowledge by mapping
‘‘elementary’’ vocal production tasks in normal subjects
in comparison with the same tasks elicited in aphasic
patients with known left hemisphere lesions. The purpose of this study was to utilize functional neuroimaging
with activation paradigms to (1) map brain structures in
normal persons for performance of well articulated,
preserved ‘‘automatic’’ speech production, as compared
with more spontaneous speech; and (2) map brain
structures in the chronic, non-fluent aphasic speaker
while they are performing these apparently different
kinds of speech tasks.
2. Materials and methods
2.1. Subjects
Fifteen subjects participated in this study after informed consent in accordance with the USC Institutional Review Board. Five patients (4 males, 1 female)
had suffered a single, unilateral stroke in perisylvian
region, resulting in non-fluent aphasia, without other
neurological or psychiatric illness. They were compared
to 10 normal-control subjects (6 males, 4 females). MRI
scans were obtained for all subjects. All subjects were
right-handed speakers of American English raised and
educated in the United States. One male normal-control
subject was eliminated due to inconsistencies in task
performance, leaving 9 subjects in the normal-control
group (see Table 1). Ages of patients ranged from 48 to
68, mean of 51.0; age range of the control subjects was
Table 1
Patient and normal-control data
Patient
Age
Ed
RW
RG
KM
LR
AC
Pt means
NC means
48
51
68
57
66
51.0
51.7
14
16
12
16
18
15.2
16.9
TPO
2.7
9
4.6
1.3
3
4.1
—
Diagnosis
AQ
Broca
Broca
Anomic
Anomic
TSA
89.7
49.9
92.6
77.2
54.3
—
—
—
—
TPO, time post-onset of injury, in years. AQ is Aphasia Quotient, a
measure of severity, from the Western Aphasia Battery. TSA, transcortical sensory aphasia. Patient and normal-control means for age
and education are given.
248
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
31–68, mean ¼ 51.7; patientsÕ education ranged from 16
to 23 years with a mean education of 15.2 years. For
normal subjects, education ranged from 16 to 23 years,
with a mean of 16.9 years. The study groups did not
differ significantly in age or education. Normal subjects
had no psychiatric, neurological, or medical disease
factors that might interfere with their status as normalcontrol subjects. None were using recreational drugs or
were on prescribed psychotropic medications. All patients had chronic (stabilized) aphasia: the time postonset of injury was between 1.3 and 9 years (mean ¼ 4.1
years). All patients were evaluated by a neurologist and
a speech pathologist. All had a diagnosis of aphasia,
distributed as follows: Broca (2), anomia (2), and
transcortical sensory aphasia (TSA) (1); all five presented clinically with superior counting over spontaneous speech, confrontation naming, or word production
ability. All were extensively evaluated by a speech-language pathologist using the Western Aphasia Battery
(Kertesz, 1982). Their scores, including the Aphasia
Quotient (AQ), a measure of aphasia severity, are shown
in Table 1.
2.2. Neurobehavioral tasks
There were three speech tasks and one silent control
task. The three types of speech elicited from all subjects
were: (1) semantically meaningful speech (words generated by recitation of animal names), (2) serial speech
(counting), and (3) non-linguistic vocalizations that engage phonation and buccal–facial movements. The latter
task required the patient to make velar sounds similar to
a ‘‘gargle,’’ lip movements involving repetitive closure
(as the ‘‘brrrr’’ sound), and palatal–nasal sounds similar
to snoring. Animal name generation was induced by
saying to the patient, ‘‘Name as many animals as you
can: think of the ocean, the zoo, the jungle, pets, the
farm, field; one example is Ôdog.’’Õ This task was selected
because it is frequently used as a measure of speech
fluency, it requires no visual input, it is an easy task to
explain to subjects, and it engages semantic categorization and word retrieval, both classic language functions.
Counting from 1 to 10 was chosen because it is the most
universally observed manifestation of preserved ‘‘automatic’’ or ‘‘non-propositional’’ speech; the majority of
aphasic patients are able to do this task; and it is simply
explained to the subject. Non-verbal vocalizations were
selected to provide a motor task similar to that used in
speech but without linguistic phonological or semantic
content. A rationale for use of this task comes from
results described in the study by McAdam and Whitaker
(1971a, 1971b), who compared verbal production
(the syllables /ka, pa/) with non-verbal vocalizations
(coughing, spitting gestures) using the auditory evoked
response paradigm, and found that only the verbal
vocalizations engaged the left hemisphere uniquely. All
three activation tasks were audiotape recorded for later
transcription and analysis.
The timing of tasks administered to each subject,
patient and normal-control, were as follows:
90 s counting 1–10 repeatedly
10-min interval
90 s naming animals
10-min interval
90 s alternating, random non-linguistic vocalizations
Eyes were open and ambient sounds were permitted
to occur throughout the activated state. In addition, a
‘‘generic’’ control condition was presented as follows:
90 s control (eyes open, ambient sounds, and no stimulation)
10-min interval
Each set (three activated scans, one resting scan) was
presented three times, resulting in a total of 12 scans.
Conditions were pseudorandomized (a different order
was given) to eliminate potential problems associated
with fixed order effects.
Verbal responses for all tasks produced by each of the
subjects were recorded on audio tape for later transcription and analysis by taping a microphone approximately 5 in. from the patientÕs mouth.
2.3. Pet brain imaging studies
Positron emission tomography (PET) using radiolabeled ð15 OÞ water provides a measure of local cerebral
blood flow, in turn reflecting local functional activation
within the human brain. 15 O PET scans were acquired
using a modified autoradiographic method (Herscovitch, Markham, & Raichle, 1983; Raichle, Martin, &
Herscovitch, 1983). For each scan, a bolus of 35 mCi of
H2 O15 was injected intravenously at the start of scanning and the speech or control task. A 90-s scan was
acquired and reconstructed using calculated attenuation
correction, with boundaries derived from the emission
scan sinogram. Arterial blood samples were not obtained. Images of radioactive counts were used to estimate relative cerebral blood flow as described previously
(Fox, Mintun, Raichle, & Herscovitch, 1984; Mazziotta
et al., 1985). PET images of rCBF were acquired with
the Siemens 953/A tomograph, which collects 31 contiguous planes covering a 105 mm field of view. The
axial resolution after reconstruction with a Hann .5 filter
was 4.3 mm at full width half maximum (FWHM) and
the trans-axial resolution was 5.5 mm FWHM as measured with a line source. The tomograph was oriented
10° steeper than the canthomeatal line to include all of
the frontal lobe, parietal lobe, and most of the cerebellum. MRI anatomic images were obtained with a GE
Signa 1.5 T device. A 3-D volumetric gradient echo
(SPGR) image of 124 contiguous slices (voxel size ¼
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
:82 :82 1:4 mm) was obtained using the sequence:
TE ¼ 5, TR ¼ 21, (flip angle ¼ 45 °). This sequence yields
excellent anatomic detail and clearly differentiates gray
and white matter.
2.4. Image analysis
All PET rCBF images were aligned to a common
stereotaxic reference frame. This stereotaxic transformation is accomplished in three steps. First, a within
subject alignment of PET scans is performed using an
automated registration algorithm (Woods, Grafton,
Holmes, Cherry, & Mazziotta, 1998a). A mean image of
the registered and resliced images is then calculated for
each subject. In the second step, the mean PET image
from each individual is co-registered to the same subjectÕs 3-D volumetric MRI scan (when available) using
another automated algorithm (Woods et al., 1998a). In
the third step, MRI scans from the different individuals
in each group (i.e., patient or control) are coregistered to
one of the subjects in that group who is centered in the
Talairach coordinate reference space by using a stepwise
transformation beginning with an affine transformation
and followed by a higher order polynomial fit with
progressively increasing degrees of freedom (Talairach
& Tournoux, 1988; Woods, Grafton, Watson, Sicotte, &
Mazziotta, 1998b). Once the MRI scans were coregistered, the same transformation matrix was applied to
the appropriate PET images. To reduce the errors secondary to repeatedly reslicing and interpolating each of
the PET images, all of the sequential reslice matrices for
each scan were combined and a single transformation
from each of the raw PET scans to the final image format in stereotaxic space was calculated.
2.5. Image analysis—statistical tests
To enhance signal detection after stereotaxic coregistration, PET rCBF images were smoothed to a final
isotropic resolution of 15 mm full width half maximum,
as verified with a line source. Because we were not interested in practice or time effects, the three repetitions
for each task were averaged prior to subsequent analysis.
The tasks used in this experiment do not necessarily
form a nested set in terms of cognitive subtraction.
Therefore, we used a multivariate data analysis method
that characterizes the variation across all four tasks and
corresponding brain activity. The method of partial least
squares (PLS) analysis was used to analyze the PET with
respect to task effects. A detailed mathematical description of this method has been described previously
(McIntosh, Bookstein, Haxby, & Grady, 1996). PLS
was chosen as the primary analysis method as it is optimized to detect relations between experimental design
factors and activity throughout the entire brain. The
design-brain PLS analysis was conducted to determine
249
how the differences in the speech tasks and control
conditions were expressed in the brain. To do this, the
cross-block correlation between orthonormal design
contrasts and each voxel of the image data set was calculated. Singular value decomposition (SVD) was used
to decompose the cross-block correlation matrix into
orthogonal pairs of singular vectors or latent variables
(LVs), which account for the covariance in the matrix in
decreasing order of importance. The vector pairs reflect
a symmetric relationship between those components of
the experimental design most related to brain activity on
one hand, and the optimal pattern of image-wide activity related to the identified design components on the
other. This second vector, termed a singular image, can
be displayed in image space. The numerical weights
within this image, called saliences, identify the collection
of voxels that as a group are most related to the design
effects expressed in the LV. The PLS analysis, in another
key result, yields brain scores for each latent variable,
which are analogous to factor scores in classic factor
analysis. Brain scores indicate how strongly individual
subjects express the patterns on the latent variable. The
scores are the minor product of the subjectÕs within scan
rCBF and the singular image on a particular LV. The
brain scores are plotted by scan to show the subject
variation in the singular images across scans. Design
scores are similarly computed by the minor product of
salience for contrasts and the contrasts themselves. This
yields a new set of contrasts that optimally code for the
effects represented in the singular image.
In this study, separate design-brain PLS analyses
were run for the normal and patient populations. Then,
a group-design-brain PLS was performed to identify
brain areas with different profiles of activation between
the two populations.
Statistical assessment of PLS was performed in two
ways. First, we tested whether the pattern of effects
represented in each of the LVs is sufficiently strong in a
statistical sense. To do this we computed the squared
multiple correlation (R2 ) from the regression of brain
scores on the design contrasts. Significance of the R2 was
assessed by means of a permutation test, using 500
permutations. Since the brain scores are derived in a
single analytic step, it is not necessary to correct for
multiple comparisons as is done for univariate image
analysis. Secondly, to determine the stability of the
saliences identified on the LVs, the standard errors of
the salience were estimated through 100 bootstrap
samples. A salience whose value depends greatly on
which subjects are in the sample is less precise than one
that remains stable regardless of the sample chosen
(Cabeza et al., 1997). The figures and tables identify all
maxima where the salience was greater than twice the
standard error. (MATLAB and C-code for PLS is
available through anonymous FTP at ftp.rotman-baycrest.on.ca/pub/randy/pls.)
250
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
To further substantiate the interpretation of the singular images, univariate analysis were performed using
the general linear model and analysis of variance. Differences in activity across scans were assessed using a
pixel by pixel ANOVA with weighted linear contrasts.
Contrast weights were defined by the design scores
generated with PLS. A significance threshold of
p < :005, uncorrected for multiple comparisons was
used. The statistical significance of the ANOVA should
be considered as descriptive only. The significance of the
LV structure comes from the permutation tests and
bootstrap (McIntosh, Lobaugh, Cabeza, Bookstein, &
Houle, 1998). Note that PLS and the univariate approaches are complementary: PLS provides an assessment of the contribution of distributed activity patterns
to the distinction between tasks and/or groups, while the
univariate analyses acts as an assessment of the importance of a given region within this larger pattern.
(Kempler, Teng, Dick, Taussig, & Davis, 1998), likely
due to the difference in mean education of the groups
(10.3 in the Kempler et al. study; 16.9 in the current
study). In that study, education was found to significantly influence animal name production (see also
Lezak, 1995).
The normal and patient study groups reported here
did not differ in education. The groups did not differ in
production of non-verbal vocalizations (255.7 vocalizations for normal subjects vs. 220.0 for patients)
(p ¼ .042) or counting (78.7 vs. 75.3 numbers) (p ¼ .40).
3.2. Imaging
Originally, data from normal and aphasic subjects
were analyzed using subtraction methodology (Van
Lancker & Grafton, 1999a, 1999b). These preliminary
3. Results
3.1. Behavioral results
Performance on the three activation tasks (naming,
vocalizations, and counting) was analyzed for all study
subjects. Quantification of vocalizations and counting
was straightforward for both study groups. For spontaneously produced animal names, repeated items, neologisms, unintelligible utterances, and non-animal
words were noted but not included in this final tabulation. Fig. 1 shows performance for naming, non-verbal
vocalizations, and counting in normal-control and patient groups; the behavioral data were compared using a
studentÕs t test. Normal-control subjects produced an
average (over three sets) of 37.8 animal names in 90 s,
compared to 11.6 animal names produced by aphasic
patients, a statistically significant difference (p < :0001).
Performance by normal subjects in this study is
higher than that reported in a recent normative study
Fig. 1. Behavioral data for normal subjects and patients.
Fig. 2. Normal vocalization. (A) Design-brain scores from PLS in
normal subjects. Weighting factors for the first latent variable after
permutation testing reveal differences of naming and vocalizing versus
counting and silence. (B) Brain saliences associated with the first significant latent variable with a ratio of salience to standard error greater
than 2 after bootstrap testing by PLS analysis are shown in color,
superimposed on the group mean MRI atlas. The results show where
naming and vocalizing > counting and control. The contours show
areas that are significant by univariate analysis using an uncorrected
threshold of p < :005.
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
analyses indicated that naming engaged language areas
in the left cerebral hemisphere, while counting did not.
However, these tasks, being linguistic (naming),
‘‘quasilinguistic’’ (counting), and non-linguistic (vocalization), cannot be conceptualized as enfolded subcomponents of each other, and therefore subtraction
methodology is especially inappropriate for analysis in
this case (Friston et al., 1996; Jennings, McIntosh, Kapur, Tulving, & Houle, 1997; Sidtis, Strother, Anderson,
& Rottenberg, 1999). Instead, partial least squares
(PLS) analysis was selected for use as an analytic tool
(McIntosh et al., 1998). This technique not only allows
activity across the entire brain to be examined, but also
avoids direct subtraction of brain states corresponding
to task conditions. The starting point of the PLS analysis is the construction of four arbitrary orthonormal
(uncorrelated) contrasts, representing critical distinctions among the four ‘‘tasks.’’
3.3. Functional localization—normal subjects
The design-brain PLS identified three significant
latent variables. The first of these identified design scores
shown in Fig. 2A. The brain scores associated with this
latent variable identify areas where naming and vocalizing are different from counting or silent control. This
LV was strongly significant ([put R2 values here if you
have them, if not put% of covariance i.e., 55% of designbrain variance] p < :0001) after permutation testing and
251
accounted for 55% of the design-brain variance. Areas
were positive brain salience scores contributed to this
effect (where naming and vocalizing > counting or control) are shown in Fig. 2B. Stability of the salience was
determined by boot-strap testing. Only sites where the
ratio of to standard error was >2 are shown and summarized in Table 2. The most prominent cortical areas
are located in bilateral precentral gyrus (mouth motor
areas), adjacent precentral sulcus (premotor cortex) and
in bilateral superior temporal gyrus including contiguous auditory cortex. Left cortical areas predominate
over right. In essence, these areas encompass known
cortical areas involved in speech production and phonological processes. The anterior cerebellum was also
identified, and subcortical activation was limited to the
left caudate. All of these sites were also significant by
standard univariate analysis, identified by the white
contour lines in Fig. 2B. The results, with the Talairach
coordinates, are presented in Table 2.
The second latent variable identified design effects
where the naming task was different from the other
tasks, as shown in Fig. 3A. This latent variable was also
strongly significant on permutation testing ðp < :0001Þ
and accounted for 31% of the design-brain variance.
Brain salience contributing positively and with stability
by boot-strap analysis are shown in Fig. 3B. The most
prominent cortical area is the left inferior and middle
frontal gyrus and inferior frontal sulcus, involving
BrodmannÕs areas 44, 45, and 6. As can be seen from
Table 2
Speech-associated regions in normal subjects (p ¼ .005)
Anatomic location (Brodmann area)
Talairach coordinates (mm)
x
PLS
Univariate
y
z
SE-max
T-max
First latent variable (naming and vocalization > control and naming)
Cerebellum
Bilateral anterior cerebellum
)2
)54
)16
8.61
5.57
L neocortical
Bilateral dorsal frontal gyrus (6)—SMA
L central sulcus (4)
L frontal insula
L inferior frontal gyrus (44)—‘‘BrocaÕs area’’
L middle frontal gyrus (9)
L middle temporal gyrus (37)
L postcentral gyrus (40)
L superior temporal gyrus (42)
8
)56
)28
)47
)40
)61
)53
)57
7
)3
22
13
40
)64
)21
)24
51
39
18
14
33
2
39
6
4.52
9.13
5.04
6.78
4.39
4.13
7.17
4.74
4.41
5.02
3.08
4.76
3.63
2.82
3.89
3.41
51
48
54
)6
)21
)32
39
42
9
4.48
6.30
7.35
4.15
3.74
4.39
)26
)20
26
3.09
2.84
R neocortical
R central sulcus (4)
R postcentral gyrus (40)
R superior temporal gyrus (42)
Non-neocortical-subcortical
L caudate
The table lists brain areas with a significant score associated with the first latent variable of the PLS analysis, with design scores corresponding to
linear contrasts where naming and vocalization > counting and control. PLS SE-max is the maximum ratio of salience to standard error for the brain
score at this location by book-strap testing. T-max and associated p-value are the results of univariate ANOVA at this location, using the design
scores from PLS as the weighting factor.
252
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
poral gyrus, and insula. Left hemisphere areas include
the precentral gyrus and superior temporal gyrus. Subcortical sites were recruited in larger proportion than in
either of the other LVs. No brain saliences were associated with thalamus or cerebellum. These sites are
summarized in Table 4.
3.4. Functional localization—patients
Fig. 3. Normal naming. (A) Design-brain scores from PLS in normal
subjects. Weighting factors for the second latent variable after permutation testing reveal differences of naming versus vocalizing,
counting, and silence. (B) Brain saliences associated with the second
significant latent variable with a ratio of salience to standard error
greater than 2 after bootstrap testing by PLS analysis are shown in
color, superimposed on the group mean MRI atlas. The results show
where naming > vocalizing, counting, and control. The contours show
areas that are significant by univariate analysis using an uncorrected
threshold of p < :005.
Table 3, cortical sites, especially those on the left, predominated, including the bilateral superior temporal
sulcus, bilateral anterior cingulate cortex and left intraparietal sulcus. The process of naming also recruited
the left thalamus and identified extensive sites in the
anterior cerebellum. The Talairach coordinates of these
sites are described in Table 3.
The third latent variable identified design effects
where the counting task was different from the other
three tasks, as shown in Fig. 4A. This latent variable was
weakly significant on permutation testing ðp < :1Þ and
accounted for 14% of the variance between design and
brain activity. The cortical brain salience associated with
this latent variable is shown in Fig. 4B. The results show
a widely distributed set of small cortical areas that are
recruited during counting compared to the other tasks.
Sites include multiple right hemispheric areas such as
inferior frontal gyrus, precentral gyrus, angular/supramarginal gyrus, middle temporal sulcus, superior tem-
MRI scans showing lesions for each of five individual
patients are shown in Fig. 5. The design-brain PLS for
the patient population identified two significant latent
variables. The first of these identified design scores
shown in Fig. 6A. The brain scores associated with this
latent variable were similar to those in normal subjects
and identified areas where naming and vocalizing are
different from counting or silent control. This LV was
significant ðp < :008Þ after permutation testing and accounted for 54% of the design-brain variance. Areas
were positive brain salience scores contributed to this
effect (where naming and vocalizing > counting or control) are shown in Fig. 6B, superimposed on a composite
image generated from all five patient MRI scans. Significant results by univariate testing are shown as contour lines. The most prominent cortical areas are located
in bilateral precentral gyrus (mouth motor areas) and
adjacent precentral sulcus (premotor cortex). There is
also recruitment of bilateral cerebellum and dorsal
frontal cortex, as in normal subjects. Unlike the results
in normal subjects, the PLS analysis did not identify
recruitment of left superior temporal gyrus. Locations of
the brain saliences are summarized in Table 5.
As in the normal population the second latent variable in the patient group identified design effects where
the naming task was different than the other tasks, as
shown in Fig. 7A. This latent variable was also significant on permutation testing ðp < :0001Þ and accounted
for 35% of the design-brain variance. Areas where brain
salience scores contributed positively to this effect are
shown in Fig. 7B. The saliences are very weak and
identify only a small subset of areas seen in the normal
population. The patients recruit the left middle and
dorsal frontal gyri, as do normal subjects. Unlike normal subjects they also recruit the right hippocampus,
right parieto-occipital fissure and left posterior parietal
cortex (Table 6).
3.5. Functional localization—normal-patient interactions
A group-design-brain PLS analysis was performed to
identify potentially significant interactions between task
and population. The analysis identified one significant
latent variable after permutation testing ðp < :0001Þ. As
shown in Fig. 8A, the design scores associated with this
latent variable identified an effect such that the difference
of naming and counting versus vocalizing and silent
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
253
Table 3
Naming-associated regions in normal subjects (p ¼ .005)
Anatomic location (Brodmann area)
Talairach coordinates (mm)
PLS
Univariate
y
z
SE-max
T-max
Second latent variable (naming > control, vocalization, and counting)
Cerebellum
L anterior cerebellum
)11
Bilateral anterior cerebellum
1
)45
)56
)13
)6
4.46
4.46
3.06
4.78
L
L
L
L
L
L
L
L
x
neocortical
inferior frontal gyrus (47)
inferior temporal gyrus (37)
inferior frontal gyrus (44, 45)
inferior frontal gyrus and frontal operculum (44, 45)
precentral sulcus (6) ‘‘premotor cortex’’
rostral dorsal frontal gyrus (10)
superior frontal gyrus (9)
)39
)55
)53
)36
)40
)8
)19
31
)47
23
21
2
66
51
)12
)11
17
8
54
)4
27
3.87
6.66
6.20
6.46
4.92
3.21
4.33
3.06
4.06
5.57
4.04
3.58
3.19
3.04
R neocortical
R dorsal frontal gyrus (32)
R dorsal frontal gyrus (6, 8)
R middle frontal gyrus (9)
R middle temporal gyrus (21)
R superior temporal (22) and inferior frontal (47) gyri
11
)7
54
67
55
29
24
20
)36
5
35
46
32
)6
)5
4.66
8.36
5.80
3.87
3.90
3.50
4.57
3.93
3.15
3.10
)29
)11
25
)38
)22
7
)9
14
3
7.64
5.31
4.13
3.89
3.58
4.43
Non-neocortical-subcortical
L hippocampus
L thalamus
R rostral putamen
The table lists brain areas with a significant score associated with the second latent variable of the PLS analysis, with design scores corresponding
to linear contrasts where naming > vocalization, counting, and control. PLS SE-max is the maximum ratio of salience to standard error for the brain
score at this location by book-strap testing. T-max and associated p-value are the results of univariate ANOVA at this location, using the design
scores from PLS as the weighting factor.
control was different for normal subjects and patients.
Fig. 8B shows areas of positive saliences where the
contrast (naming, counting > vocalizing, and control) is
greater in normals than patients. It demonstrates that
normal subjects are able to recruit a widely distributed
set of cortical areas throughout the left frontal parietal
and temporal cortex during naming and vocalizing, and
that they perform the tasks differently than the patients.
4. Discussion
In normal subjects, different patterns of cerebral activation were associated with three speech production
tasks: word generation (‘‘naming’’), (non-linguistic) vocalization, and counting (1–10). Using a partial least
squares analysis to detect how differences in the speech
tasks were expressed in the brain, distinct profiles of
activation were identified. Non-verbal vocalization and
naming were associated with activation in areas previously associated with speech production. Bilateral sites
include motor cortex in the mouth area, the supplementary motor area and the anterior cerebellum. All of
these areas are associated with articulatory motor control. The result is consistent with the notion that naming
and vocalizing are more engaging of articulatory pro-
cesses than counting or silence. The latent variable also
identified activation in bilateral superior temporal gyri.
This area is known to be involved in auditory processing, with increasing activity for more complex auditory
stimuli. There was more activity in this area for naming
or vocalizing compared to the counting and silence
tasks. In this profile, subcortical saliences were minimal,
and right sites appeared as only 25% of the total identified sites.
The second latent variable identified areas where
naming was different than the other three tasks. The
strongest contribution for this effect was located in the
left middle and inferior frontal gyri, also known as
BrocaÕs area. It is emphasized that the right middle
frontal lobe, and bilateral anterior cingulate and superior temporal sulci as well as left hippocampus and left
parietal lobule were also involved. Categorical noun
generation requires a multitude of cortical systems for
recall, attention and word production.
The third latent variable identified areas contributing
to the production of automatic speech (counting) to a
greater degree than the other tasks. What is striking
about the result in normals is the multitude of both
cortical and subcortical sites within both hemispheres
that contribute to this automatic task (Fig. 9). For this
variable, increased subcortical sites was seen, and right
254
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
Fig. 4. Normal counting. (A) Design-brain scores from PLS in normal
subjects. Weighting factors for the third latent variable after permutation testing reveal differences of counting versus naming, vocalizing,
and silence. (B) Brain saliences associated with the third significant
latent variable with a ratio of salience to standard error greater than 2
after bootstrap testing by PLS analysis are shown in color, superimposed on the group mean MRI atlas. The results show where counting
versus naming, vocalizing and silence. The contours show areas that
are significant by univariate analysis using an uncorrected threshold of
p < :005.
cortical involvement predominated over left. Presuming
that there is functional redundancy within this distributed set of areas, it is not surprising that patients with
unilateral cortical lesions causing aphasia may have
preservation of automatic speech. However, whether
specific sites selectively and uniquely subserve counting
cannot yet be determined.
The first latent variable derived from the aphasic
patient results indicated that intact naming and nonverbal vocalization were associated with diminished
activation throughout right hemisphere structures normally involved in articulation as well as reduced activity
in left sided structures outside of the stroke lesion. There
were no novel areas of recruitment to suggest functional
reorganization of articulatory areas. The inter-group
PLS also identified differences of areas involved in articulation (Fig. 8). In the direct comparison, the main
finding was greater activation throughout preserved left
frontal temporal and parietal cortex in the normal
subjects compared to patients.
The results of the second latent variable, which was
associated with naming alone, showed a preservation of
activation in left middle frontal gyrus and left mesial
frontal cortex. These areas, outside of the lesion, are
normally recruited during naming. The patients also
recruited several novel sites not active in the normal
subjects, including the right hippocampus and mesial
parieto-occipital cortex. We can not exclude the possibility that recruitment of these areas is due to task difficulty or effort. Nevertheless it is noteworthy that the
mesial parieto-occipital cortex is known to be involved
in visualization. It is possible that the patients use a
different cognitive strategy than normals—one that involves more visual recollection than categorical noun
generation.
The third latent variable in the patients was not significant. More importantly, there was no evidence that
patients used a different set of brain areas for counting
compared to the other tasks. In contrast, normal subjects do. It might be inferred that a residual set of undamaged regions in patients are used for both automatic
and non-automatic speech. From the normal group it
appears these areas are widely distributed throughout
both cerebral hemispheres and subcortical sites.
It is useful to consider these functional imaging
studies in light of previous models of language localization derived from lesion lesions, and later from a
broader array of paradigms. The case for the left
hemisphere subserving residual utterances comes from
the older view of the left hemisphere as primary and as
the only source of linguistic output. This view is supported by the frequency of speech deficits following left
hemisphere damage and the paucity of speech deficits
following right hemisphere damage. However, a role of
the right hemisphere in linguistic competency has been
postulated on the basis of symptomatic worsening in
left-brain-injured aphasic patients after temporary right
hemisphere inactivation by intracarotid amobarbital
injection (Czopf, 1981; Kinsbourne, 1971) or permanent
damage brought about by a new stroke to the previously
intact right hemisphere (Cummings, Benson, Walsh, &
Levine, 1979; Mohr & Levine, 1979). Earlier studies of
cerebral blood flow associated right hemisphere activation with ‘‘automatic speech’’ (Ingvar, 1983; Larsen,
Skinhoj, & Lassen, 1978; Ryding, Bradvik, & Ingvar,
1987). An adult left-hemispherectomized subject, who
became profoundly aphasic following surgery, was able
to use speech formulas and swear (Smith, 1966; Smith &
Burklund, 1966; Van Lancker & Cummings, 1999).
During right-sided/LH injection in a clinical Wada
procedure (Wada & Rasmussen, 1960) observed by an
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
255
Table 4
Counting-associated regions in normal subjects (p ¼ .005)
Anatomic location (Brodmann area)
Talairach coordinates (mm)
PLS
Univariate
y
z
SE-max
T-max
Third latent variable (counting > naming, vocalization, and control)
L neocortical
L cuneate cortex (19)
)16
L dorsal frontal gyrus (11)
)10
L precentral gyrus (6), superior temporal gyrus (42)
)51
)83
22
)9
25
)13
14
4.76
3.27
3.55
3.52
3.30
2.97
R neocortical
R angular/supramarginal gyrus (39)
R inferior frontal gyrus (44)
R insula
R middle temporal sulcus (20)
R superior temporal gyrus (42, 22)
42
40
38
45
35
)49
9
1
)31
)26
32
31
2
)11
14
2.97
2.97
2.85
3.30
4.73
2.93
2.97
2.99
3.60
3.67
7
)28
14
24
11
)59
)19
2
)38
)17
9
15
)2
)4
)5
3.00
4.88
3.55
2.73
7.73
3.67
3.12
3.39
2.80
3.93
x
Non-neocortical-subcortical
L cingulate gyrus (23/31)
L putamen
R globus pallidus
R hippocampus
R subthalamic nucleus/red nucleus
The table lists brain areas with a significant score associated with the third latent variable of the PLS analysis, with design scores corresponding to
linear contrasts where counting > vocalization, naming, and control. PLS SE-max is the maximum ratio of salience to standard error for the brain
score at this location by book-strap testing. T-max and associated p-value are the results of univariate ANOVA at this location, using the design
scores from PLS as the weighting factor.
author of this study (DVS), the patient, who was unable
to respond to any aspect of the language testing, counted from 1 to 10 repeatedly. These observations, as well
Fig. 5. Localization of lesions in five stroke patients. Three-dimensional reconstruction of structural MRI studies from the five patients
demonstrating the location of left hemispheric infarcts. In subject 4 the
lesion was entirely subcortical.
as the experimental study by Graves and Landis (1985)
mentioned above, lead to the notion of the RH as substrate for some automatic speech behaviors. This assumption is supported by countless clinical observations
of preserved speech in a large array of cases of extensive
left hemisphere damage, leading to the inference that
across these many cases, the right hemisphere is likely to
subserve residual aphasic speech.
A key subcortical role in automatic speech has been
inferred from informal observations of reduced output
of such expressions in ParkinsonÕs disease, hyperactivation of expletives in TouretteÕs syndrome (Van Lancker & Cummings, 1999), and the case report (Speedie
et al., 1993) mentioned earlier describing a decrement in
production of overlearned prayers and counting following an infarct in the right basal ganglia. Two neural
systems for different kinds of cognitive processing, often
referred to as procedural and declarative and correlating
with habitual versus novel behaviors (Lounsbury, 1963;
Sinclair, 1991) have been proposed, and may be associated with subcortical and cortical structures, respectively (Lieberman, 2001; Mishkin, Malamut, &
Bachevalier, 1984; Mishkin & Petri, 1984; Robinson,
1976). Counting, as a subset of serial (automatic)
speech, might well be classed as a behavior dependent on
procedural memory, as contrasted to word production,
which requires declarative (semantic) memory in normal
subjects. In our study, more subcortical sites were
identified in the third latent variable (LV), the one associated with counting, than in either of the other two
LVs, which were associated with naming and vocalization (see Fig. 9).
256
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
Fig. 6. Patient vocalization. (A) Design-brain scores from PLS in the
five patients. Weighting factors for the first latent variable after permutation testing reveal differences of naming and vocalizing versus
counting and silence. (B) Brain saliences associated with the first significant latent variable with a ratio of salience to standard error greater
than 2 after bootstrap testing by PLS analysis are shown in color,
superimposed on the group mean MRI atlas. The results show where
naming and vocalizing > counting and control. The contours show
areas that are significant by univariate analysis using an uncorrected
threshold of p < :005.
Fig. 7. Patient naming. (A) Design-brain scores from PLS in patient
subjects. Weighting factors for the second latent variable after permutation testing reveal differences of naming versus vocalizing,
counting, and silence. (B) Brain saliences associated with the second
significant latent variable with a ratio of salience to standard error
greater than 2 after bootstrap testing by PLS analysis are shown in
color, superimposed on the group mean MRI atlas. The results show
where naming > vocalizing, counting, and control. The contours show
areas that are significant by univariate analysis using an uncorrected
threshold of p < :005.
Several types of converging evidence suggest that
counting as a speech behavior differs from other speech
behaviors. In the cortical stimulation studies of Penfield
and Roberts (1959), various kinds of naming errors as
well as ‘‘confusion of numbers while counting’’ (p. 133)
occurred in frontal, temporal, and parietal areas of the
left hemisphere. The authors predict that misnaming
and counting errors will also occur from the right
hemisphere (pp. 126–127), but no actual incidences were
reported. In these studies, number confusion was
grouped with other ‘‘dysphasic or aphasic types of responses’’ (e.g., p. 130). In a study of the effects of cortical stimulation on speech production abilities, one
patient was able to count forward from one by ones
during stimulation of the basal temporal language area,
but was unable to count by threes or say the days of the
week (Lueders et al., 1991). He repeatedly reverted to
counting by ones while attempting these other tasks.
Confrontation naming was the most affected by stimulation of the language area with reading aloud and
repetition of words also showing interruptions and paraphasias; counting alone remained fluent. Similarly, in
a stimulation study using subdural electrodes, at one site
speech arrest was observed during a naming task but
counting was not disrupted (Fried et al., 1991). These
stimulation studies support our findings that counting
and naming involve different brain structures.
A preliminary report using PET imaging indicated
differences in brain activation patterns for counting
compared with story telling (Blank, Scott, & Wise,
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
257
Table 5
Speech-associated regions in stroke patients (p ¼ .005)
Anatomic location (Brodmann area)
Talairach coordinates (mm)
PLS
Univariate
y
z
SE-max
T-max
)63
)9
4.47
5.85
0
)51
)13
2
1
)4
58
43
56
3.77
3.05
3.35
3.39
4.37
3.51
38
48
)9
)30
34
4
2.68
2.86
4.39
5.15
6
)15
3
3.53
2.84
x
First latent variable (naming and vocalizing > control and counting)
Cerebellum
Bilateral anterior cerebellum
)8
L neocortex
Bilateral dorsal frontal gyrus (6)—preSMA
L precentral gyrus (4/6)
L dorsal frontal gyrus (6)—SMA
R neocortex
R precentral gyrus (4/6)
R superior temporal sulcus (21, 22)
Non-neocortex-subcortical
R thalamus
The table lists brain areas with a significant score associated with the first latent variable of the PLS analysis, with design scores corresponding to
linear contrasts where naming and vocalization > counting and control. PLS SE-max is the maximum ratio of salience to standard error for the brain
score at this location by book-strap testing. T-max and associated p-value are the results of univariate ANOVA at this location, using the design
scores from PLS as the weighting factor.
Table 6
Naming-associated regions in stroke patients (p ¼ .005)
Anatomic location (Brodmann area)
Talairach coordinates (mm)
PLS
Univariate
SE-max
T-max
p-Value
x
y
z
)11
)32
)20
29
37
)57
46
35
56
2.27
2.41
3.17
3.13
3.24
1.24
.005
.005
NS
R neocortical
R parieto-occipital fissure
9
)65
17
2.46
5.11
.005
Non-neocortical-subcortical
R hippocampus
26
)10
)9
2.69
3.17
.005
L
L
L
L
neocortical
dorsal frontal gyrus (6/8)
middle frontal gyrus (9)
superor parietal lobule (7)
The table lists brain areas with a significant score associated with the second latent variable of the PLS analysis, with design scores corresponding
to linear contrasts where naming > vocalization, counting, and control. PLS SE-max is the maximum ratio of salience to standard error for the brain
score at this location by book-strap testing. T-max and associated p-value are the results of univariate ANOVA at this location, using the design
scores from PLS as the weighting factor.
2001). Another study using PET imaging employed two
speech tasks traditionally considered to be ‘‘automatic,’’
a serial task (months of the year) and a well rehearsed,
memorized text (the Pledge of Allegiance) (Bookheimer,
Zeffiro, Blaxton, Gaillard, & Theodore, 2000), compared to tongue movements and consonant–vowel syllable production. Continuous production of the Pledge
of Allegiance showed activation in traditional language
areas; reciting the months of the year selectively engaged, of language areas, Brodmann areas 44 and 22.
These observations do not cast light on counting, which
has been the most widely used task in cortical mapping,
and is the most frequently observed preserved automatic
speech behavior.
Clinical observations in chronic aphasia suggest that
nearly all language-afflicted patients can count to ten; in
contrast, a smaller number can recite other serial lists
(e.g., the days of the week and the alphabet to G are
performed more readily than months of the year).
Ability to produce well established, longer discourse
units, such as prayers and song lyrics, including the
Pledge of Allegiance, is more variable and appears less
frequently. It has earlier been proposed that speech
competence is usefully viewed on a continuum from
wholly novel (newly created) utterances to reflexive
cries, with serial lists, memorized speech, interactional
speech formulas, idioms, and other categories taking
places along this continuum, according to the properties
of each (Van Lancker, 1988). Brain function underlying
these categories, as well as important phenomenological
differences between them, remain to be studied and
understood.
258
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
Fig. 8. Brain areas where there is a significant interaction between
design-group-brain activity after PLS analysis. In panel A, areas in
blue correspond to regions where normal subjects show greater activity
during naming and counting compared to vocalizing or control relative
to the patient group. These interactions are characterized in panel B.
studies have associated semantic and phonological
processing with left inferior prefrontal cortex (Demonet
et al., 1992; Desmond et al., 1995; Poldrack et al., 1999;
Thompson-Schill, DÕEsposito, Aguirre, & Farah, 1997;
but see Price, Moore, Humphreys, & Wise, 1997). In the
present study, vocalization and naming patterned together, and the brain-group-profiles identified were primarily left prefrontal sites. Although the vocalization
task was intended to be ‘‘non-linguistic,’’ in actuality,
subjects produced sequences of phonological sounds.
In summary, counting was not associated with coherently patterned brain areas in normal subjects;
counting did not yield a significant set of brain sites in
patients; and counting alone failed to distinguish between patient and normal groups. Counting patterned
more consistently with the control (rest) task, suggesting
that counting does not fall in the category with phonological or semantic processing. These results suggest
that counting is not an optimal task for intraoperative
cortical speech mapping. Counting may also not be
optimal as a reference state or baseline in functional
brain imaging studies (Hutchinson et al., 1999; Pihlajam€aki et al., 2000). In agreement with the clinical observations on persons with aphasia, for whom counting
is easier than word production, and who have various
degrees of left hemisphere cortical damage, counting
was identified with a greater array of subcortical and
RH cerebral sites than word production in the present
study. Functional mapping for surgical planning is likely
to be more successful when the essential differences between propositional and non-propositional speech are
recognized.
Uncited reference
(Graves, Landis, & Simpson, 1985).
Acknowledgments
Fig. 9. Schematic chart of relative numbers of brain sites associated
with three latent variables identified as naming-vocalization, naming,
and counting in normal subjects.
In the study reported here, an unexpected result was
that the task designated as ‘‘non-verbal vocalizations’’
and the word generation task (or ‘‘naming’’) patterned
together in both study groups for the greatest proportion of variance (the first latent variable). These results
are often difficult to compare with findings from other
imaging studies, because of differences in design, subject
population, and task demands, such as use of silent
speech (e.g., Friedman et al., 1998). However, several
Supported in part by a McDonnell-Pew Foundation
award to Diana Van Lancker and Scott Grafton and
PHS awards NS33504 and NS37470 to Scott Grafton.
Varsha Pilbrow assisted with manuscript preparation.
We appreciate the comments of John J. Sidtis on an
earlier version of this article.
References
Albert, M. L., & Helm-Estabrooks, N. (1988). Diagnosis and
treatment of aphasia. Part II. Journal of the American Medical
Association, 259, 1205–1210.
Benson, D. F. (1979). Aphasia, alexia, and agraphia. New York:
Churchill Livingston.
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
Berger, M. S., & Ojemann, G. A. (1992). Intraoperative brain mapping
techniques in neuro-oncology. Stereotactic and functional neurosurgery, 58, 153–161.
Blank, C., Scott, S., & Wise, R. (2001). Neural systems involved in
propositional and non-propositional speech. In 7th annual meeting
of the organization for human brain mapping, Brighton, UK, June
10–14 (Abstract # 11192).
Blanken, G. (1991). The functional basis of speech automatisms
(recurring utterances). Aphasiology, 5, 103–127.
Blanken, G., Wallesch, E.-W., & Papagno, C. (1990). Dissociations of
language functions in aphasics with speech automatisms (recurring
utterances). Cortex, 26, 41–63.
Broca, P. (1865). S^
ur le siege de la faculte du langage articule. Bulletins
de la Societe dÕAnthropologie de Paris (Paris), 6, 377–393.
Bookheimer, S. Y., Zeffiro, T. A., Blaxton, T. A., Gaillard, P. W., &
Theodore, W. H. (2000). Activation of language cortex with
automatic speech tasks. Neurology, 55, 1151–1157.
Cabeza, R., & Nyberg, L. (2000). Imaging cognition II: An empirical
review of 275 PET and fMRI studies. Journal of Cognitive
Neuroscience, 12, 1–47.
Cabeza, R., Grady, C. L., Nyberg, L., McIntosh, A. R., Tulving, E.,
Kapur, S., Jennings, J. M., Houle, S., & Craik, F. I. (1997). Agerelated differences in neural activity during memory encoding and
retrieval: A positron emission tomography study. Journal of
Neuroscience, 17, 391–400.
Celesia, G. G., Polcyn, R. E., Holden, J. E., Nickles, R. J., Koeppe, R.
A., & Gatley, S. J. (1984). Determination of regional cerebral blood
flow in patients with cerebral infarction: Use of fluoromethane
labeled with flourine 18 and positron emotion tomography.
Archives of Neurology, 41, 262–267.
Code, C. (1989). The characteristics of aphasia. London: Taylor &
Francis.
Code, C. (1982). Neurolinguistic analysis of recurrent utterance in
aphasia. Cortex, 18, 141–152.
Critchley, M. (1962). Speech and speech-loss in relation to the duality
of the brain. In V. Mountcastle (Ed.), Interhemispheric relations and
cerebral dominance (pp. 208–213). Baltimore: Johns Hopkins
University Press.
Critchley, M. (1970). Aphasiology and other aspects of language.
London: Edward Arnold.
Cummings, J., Benson, D. F., Walsh, M. J., & Levine, H. L. (1979).
Left-to-right transfer of language dominance: A case study.
Neurology, 29, 1547–1550.
€ ber die Rolle der nicht dominanten Hemisph€are in
Czopf, J. (1981). U
der Restitution der Sprache der Aphasischen. Archiven Psychiatrischen Nervenkrankheiten, 216, 162–171.
Damasio, H., Grabowski, T., Tranel, D., Hichwa, R., & Damasio, A.
(1996). A neural basis for lexical retrieval. Nature, 380(6574), 499–
505.
Demonet, J.-F., Wise, R., & Frackowiak, R. S. J. (1993). Language
functions explored in normal subjects by positron emission
tomography: A critical review. Human Brain Mapping, 1, 39–
47.
Demonet, J.-F., Chollet, F., Ramsay, S., Cardebat, D., Nespoulous, J.L., Wise, R., Rascol, A., & Frackowiak, R. (1992). The anatomy of
phonological and semantic processing in normal subjects. Brain,
115, 1753–1768.
Desmond, J. E., Sum, J. M., Wagner, A. D., Demb, J. B., Shear, P. K.,
Glover, G. H., Gabrieli, J. D. E., & Morrell, M. J. (1995).
Functional MRI measurement of language lateralization in Wadatested patients. Brain, 18, 1411–1419.
Espir, L., & Rose, F. (1970). The basic neurology of speech. Oxford:
Blackwell.
Fiez, J. A., Raichle, M. E., Miezin, F. M., Petersen, S. E., Tallal, P., &
Katz, W. (1995). PET studies of auditory and phonological
processing: Effect of stimulus characteristics and task demands.
Journal of Cognitive Neuroscience, 7, 357–375.
259
Fox, P. T., Mintun, M. A., Raichle, M. E., & Herscovitch, P. (1984). A
non-invasive approach to quantitative functional brain mapping
with H2 15 O and positron emission tomography. Journal of Cerebral
Blood Flow Metabolism, 4, 329–333.
Fried, I., Katz, A., McCarthy, G., Sass, K. J., Williamson, P., Spencer,
S. S., & Spencer, D. D. (1991). Functional organization of human
supplementary motor cortex studied by electrical stimulation. The
Journal of Neuroscience, 11, 3656–3666.
Fried, I., Nenov, V., Ojemann, S. O., & Woods, R. (1995). Functional
MR and PET imaging of rolandic and visual cortices for
neurosurgical planning. Journal of Neurosurgery, 83, 854–861.
Friedman, L., Kenny, J. T., Wise, A. L., Wu, D., Stuve, T. A., Miller,
D. A., Jesberger, J. A., & Lewin, J. S. (1998). Brain activation
during silent word generation evaluated with functional MRI.
Brain and Language, 64, 231–256.
Friston, K. J., Frith, C. D., Liddle, P. F., & Frackowiak, R. S. J.
(1991). Investigating a network model of word generation with
positron emission tomography. Proceedings of the Royal Society of
London, 244, 101–106.
Friston, K. J., Price, C. J., Fletcher, P., Moore, C., Frackowiak, R. S.
J., & Dolan, R. J. (1996). The trouble with cognitive subtraction.
NeuroImage, 4, 97–104.
Graves, R., & Landis, T. (1985). Hemispheric control of speech
expression in aphasia. A mouth asymmetry study. Archives of
Neurology, 42(3), 249–251.
Graves, R., Landis, T., & Simpson, C. (1985). On the interpretation of
mouth asymmetry. Neuropsychologia, 23, 121–123.
Haxby, J. V., Grady, C. L., Ungerleider, L. G., & Horwitz, B. (1991).
Mapping the functional neuroanatomy of the intact human brain
with brain work imaging. Neuropsychologia, 29, 539–555.
Henderson, V. W. (1985). Lesion localization in BrocaÕs aphasia.
Implications from BrocaÕs aphasia without hemiparesis. Archives of
Neurology, 42, 1210–1212.
Henderson, V. W. (1987). Language disorders: Clinical classification
and neurovascular substrate. Bulletin of Clinical Neurosciences, 52,
70–88.
Herscovitch, P., Markham, J., & Raichle, M. E. (1983). Brain blood
flow measured with intravenous H2 15 O. I. Theory and error
analysis. Journal of Nuclear Medicine, 24, 782–789.
Howard, D., Patterson, K., Wise, R., Brown, W. D., Friston, K.,
Weiller, C., & Frackowiak, R. (1992). The cortical localization of
the lexicons. Positron emission tomography evidence. Brain, 115,
1769–1782.
Hunter, K. E., Blaxton, T. A., Bookheimer, S. Y., Figlossi, C.,
Gaillard, W. D., Grandin, C., Anyanwu, A., & Theodore, W. H.
(1999). Sup-1–sup-50 water positron emission tomography (PET)
in language localization: A study comparing positron emission
tomography visual and computerized region of interest analysis
with the Wada test. Annals of Neurology, 45, 662–665.
Hutchinson, M., Schiffer, W., Joseffer, S., Liu, A., Schlosser, K.,
Kikshit, S., Goldberg, E., & Brodie, J. D. (1999). Task-specific
deactivation patterns in functional magnetic resonance imaging.
Magnetic Resonance Imaging, 17, 1427–1436.
Ingvar, D. H. (1983). Serial aspects of language and speech related to
prefrontal cortical activity. A selective review. Human Neurobiology, 2, 177–189.
Jackson, J. H. (1878). On affections of speech from disease of the brain.
In J. Taylor (Ed.), Selected writings of John Hughlings Jackson (Vol.
2: 1878–1879/1932, pp. 155–204). London: Hodder & Stoughton,
1932 (Originally in Brain 1 (1878/1879) 304–330; Brain 2 (1879–
1880) 203–222 & 323–356; also Reprinted in Brain 38 (1915)).
Jennings, J. M., McIntosh, A. R., Kapur, S., Tulving, E., & Houle, S.
(1997). Cognitive subtractions may not add up: The interaction
between semantic processing and response mode. NeuroImage, 5,
229–239.
Karbe, H., Thiel, A., Weber-Luxenburger, G., Herholz, K., Kessler, J.,
& Heiss, W. D. (1998). Brain plasticity in poststroke aphasia: What
260
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
is the contribution of the right hemisphere? Brain and Language,
64, 215–230.
Kempler, D., Teng, E. L., Dick, M., Taussig, I. M., & Davis, D. S.
(1998). The effects of age, education and ethnicity on verbal
fluency. Journal of the International Neuropsychological Society, 4,
531–538.
Kertesz, A. (1982). The Western Aphasia Battery. New York: Grune
and Stratton.
Kinsbourne, M. (1971). The minor cerebral hemisphere as a source of
aphasic speech. Transactions of the American Neurological Association, 96, 141–145.
Kircher, T. T., Brammer, M. J., Williams, S. C., & McGuire, P. K.
(2000). Lexical retrieval during fluent speech production: An fMRI
study. NeuroReport, 11, 4093–4096.
Klein, D., Olivier, A., Milner, B., Zatorre, R. J., Johnsrude, I. M., &
Evans, A. C. (1997). Obligatory role of the LIFG in synonym
generation: Evidence from PET and cortical stimulation. NeuroReport, 8, 3275–3279.
Lange, N., Strother, S., Anderson, J., Nielsen, F., Holmes, A.,
Kolenda, T., Savoy, R., & Hansen, L. (1999). Plurality and
resemblance in fMRI data analysis. NeuroImage, 10, 282–303.
Larsen, B., Skinhoj, E., & Lassen, H. A. (1978). Variations in regional
cortical blood flow in the right and left hemispheres during
automatic speech. Brain, 10, 193–200.
Leblanc, R., Meyer, E., Bub, D., Zatorre, R., & Evans, A. (1992).
Language localization with activation positron emission tomography scanning. Neurosurgery, 31, 369–373.
Lebrun, Y., & Leleux, C. (1993). The effects of electrostimulation and
of resective and stereotactic surgery on language and speech. Acta
Neurchirurgica Supplementum, 56, 40–51.
Lieberman, P. (2001). Human language and our reptilian brain.
Perspectives in Biology and Medicine, 44, 32–51.
Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New
York: Oxford University Press.
Lounsbury, F. G. (1963). Linguistics and psychology. In S. Koch
(Ed.), Psychology: Study of a science (pp. 553–582). New York:
McGraw-Hill.
Lueders, H., Lesser, R. P., Hahn, J., Dinner, D. S., Morris, H. H.,
Wyllie, E., & Godoy, J. (1991). Basal temporal language area.
Brain, 114, 743–754.
Lum, C. C., & Ellis, A. W. (1994). Is ÔnonpropositionalÕ speech
preserved in aphasia? Brain and Language, 46, 368–391.
Mateer, C. A. (1983). Localization of language and visuospatial
functions by electrical stimulation. In A. Kertesz (Ed.), Localization in neuropsychology (pp. 153–183). New York: Academic Press.
Mazziotta, J. C., Huang, S. C., Phelps, M. E., Carson, R. E.,
MacDonald, N. S., & Mahoney, K. (1985). A noninvasive positron
computed tomography technique using oxygen-15-labeled water
for the evaluation of neurobehavioral task batteries. Journal of
Cerebral Blood Flow Metabolism, 5, 70–78.
McAdam, D. W., & Whitaker, H. A. (1971a). Language production:
Electroencephalographic localization in the normal human brain.
Science, 172, 499–502.
McAdam, D. W., & Whitaker, H. A. (1971b). Language production:
Electrocortical localization of language production. Science, 174,
1359–1361.
McIntosh, A. R., Bookstein, F. L., Haxby, J. V., & Grady, C. L.
(1996). Spatial pattern analysis of functional brain images using
partial least squares. NeuroImage, 3, 143–157.
McIntosh, A. R., Lobaugh, N. J., Cabeza, R., Bookstein, F. L., &
Houle, S. (1998). Convergence of neural systems processing
stimulus associations and coordinating motor responses. Cerebral
Cortex, 8(7), 649–659.
Metter, E. J., Hanson, W. R., Jackson, C. A., Kempler, D., Van
Lancker, D., Mazziotta, J. C., & Phelps, M. E. (1990). Temporal
parietal cortex in aphasia: Evidence from positron emission
tomography. Archives of Neurology, 47, 1235–1238.
Mishkin, M., Malamut, B., & Bachevalier, J. (1984). Memories and
habits: Two neural systems. In G. Lynch, J. L. McGaugh, & N. M.
Weinberger (Eds.), Neurobiology of learning and memory (pp. 65–
77). New York: Guilford Press.
Mishkin, M., & Petri, H. L. (1984). Memories and habits: Some
implications for the analysis of learning and retention. In L. R.
Squire, & N. Butters (Eds.), Neuropsychology of memory (pp. 287–
296). New York: Guilford Press.
Mohr, J. P., & Levine, D. N. (1979). Language after bilateral cerebral
infarction: Role of the minor hemisphere in speech. Neurology, 29,
927–938.
Ojemann, G. A. (1983). Brain organization for language from the
perspective of electrical stimulation mapping. The Brain and
Behavioral Sciences, 6, 189–230.
Ojemann, G. A. (1994). Cortical stimulation and recording in
language. In A. Kertesz (Ed.), Localization and neuroimaging in
neuropsychology (pp. 35–55). San Diego: Academic Press.
Ojemann, G. A. (1995). Integrating neuropsychological and neurosurgical evaluations in patients coming to epilepsy surgery. Workshop
presented at the international neuropsychological society meetings,
Seattle, Washington, February 1995 (personal communication.
From C. Dodrill, & G. Ojemann).
Paus, T., Petrides, E., Evans, A., & Meyer, E. (1993). Role of the
human anterior cingulate cortex in the control of oculomotor,
manual, and speech responses: A positron emission tomography
study. Journal of Neurophysiology, 70, 453–469.
Penfield, W., & Roberts, L. (1959). Speech and brain-mechanisms.
Atheneum, New York: Princeton University Press.
Petersen, S. E., & Fiez, J. A. (1993). The processing of single words
studied with positron emission tomography. Annual Review of
Neuroscience, 16, 509–530.
Petersen, S. E., Fox, P. T., Posner, M. I., Mintun, M., & Raichle, M.
E. (1988). Positron emission tomographic studies of the cortical
anatomy of single-word processing. Nature, 331, 585–589.
Pihlajam€aki, M., Tanila, H., H€anninen, T., K€
on€
onen, M., Laakso, M.,
Partanen, K., Soininen, H., & Aronen, H. (2000). Verbal fluency
activates the left medial temporal lobe: A functional magnetic
resonance imaging study. Annals of Neurology, 47, 470–476.
Poldrack, R. A., Wagner, A. D., Prull, M. W., Desmond, J. E., Glover,
G. H., & Gabrieli, J. D. E. (1999). Functional specialization for
semantic and phonological processing in the left inferior prefrontal
cortex. NeuroImage, 10, 15–35.
Price, C. J., Wise, J. J. S., Warburton, E. A., Moore, C. J., Howard,
D., Patterson, K., Frackowiak, R. S. J., & Friston, K. J. (1996).
Hearing and saying. The functional neuro-anatomy of auditory
word processing. Brain, 119, 919–931.
Price, C. J., Moore, C. J., Humphreys, G. W., & Wise, R. J. S. (1997).
Segregating semantic from phonological processes during reading.
Journal of Cognitive Neuroscience, 9, 727–733.
Raichle, M. E., Martin, W. R. W., & Herscovitch, P. (1983). Brain
blood flow measured with intravenous H2 15 O. II. Implementation
and validation. Journal of Nuclear Medicine, 24, 790–798.
Raichle, M. E. (1991). Memory mechanisms in the processing of
words and word-like symbols. Exploring brain functional anatomy with positron tomography. CIBA Foundation Symposium,
163, 198–217.
Robinson, B. W. (1976). Limbic influences on human speech. Annals of
the New York Academy of Sciences, 280, 761–771.
Ryding, E., Bradvik, B., & Ingvar, D. H. C. (1987). Changes of
regional cerebral blood flow measured simultaneously in the right
and left hemisphere during automatic speech and humming. Brain,
110, 1345–1358.
Sidtis, J. J. (2000). From chronograph to functional image: WhatÕs
next? Brain and Cognition, 42, 75–77.
Sidtis, J. J., Anderson, J., Strother, S., & Rottenberg, D. (1998).
Predicting performance from functional imaging data. NeuroImage, 7, S749.
D. Vanlancker-Sidtis et al. / Brain and Language 85 (2003) 245–261
Sidtis, J. J., Strother, S. C., Anderson, J. R., & Rottenberg, D. A.
(1999). Are brain functions really additive? NeuroImage, 9, 490–
496.
Sinclair, J. M. (1991). Corpus, concordance, collocation. Oxford:
Oxford University Press.
Smith, A. (1966). Speech and other functions after left (dominant)
hemispherectomy. Journal of Neurology, Neurosurgery and Psychiatry, 29, 467–471.
Smith, A., & Burklund, C. W. (1966). Dominant hemispherectomy.
Science, 153, 1280–1282.
Speedie, L. J., Wertman, E., TaÕir, J., & Heilman, K. M. (1993).
Disruption of automatic speech following a right basal ganglia
lesion. Neurology, 43, 1768–1774.
Steinmetz, H., & Seitz, R. J. (1991). Functional anatomy of language
processing: Neuroimaging and the problem of individual variability. Neuropsychologia, 29, 1149–1161.
Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the
brain. New York: Thieme Medical Publishers.
Thompson-Schill, S. L., DÕEsposito, M., Aguirre, G. K., & Farah, M.
J. (1997). Role of left inferior prefrontal cortex in retrieval of
semantic knowledge: A reevaluation. Proceedings of the National
Academy of Science of the United States of America, 94, 14792–
14797.
Tomczak, R. J., Wunderlich, A. P., Wang, Y., Braun, V., Antoniadis,
G., Goerich, J., Richter, H.-P., & Brambs, H.-J. (2000). FMRI for
preoperative neurosurgical mapping of motor cortex and language
in a clinical setting. Journal of Computer Assisted Tomography, 24,
927–934.
Vandenberge, R., Price, C., Wise, R., Josephs, O., & Frackowiak, R. S.
J. (1996). Functional anatomy of a common semantic system for
words and pictures. Nature, 383, 254–256.
Van Lancker, D. (1988). Nonpropositional speech: Neurolinguistic
studies. In A. E. Ellis (Ed.), Progress in the psychology of language:
Vol. III (pp. 49–118). Hillsdale, NJ: Lawrence Erlbaum.
Van Lancker, D. (1993). Nonpropositional speech in aphasia. In G.
Blanken, J. Dittmann, H. Grimm, J. C. Marshall, & C. W.
261
Wallesch (Eds.), Linguistic disorders and pathologies. Berlin: Walter
de Gruyter.
Van Lancker, D., & Bella, R. (1996). The relative roles of repetition
and sentence completion tasks in revealing superior speech abilities
in patients with nonfluent aphasia. Journal of the International
Neuropsychological Society, 2, 6.
Van Lancker, D., & Cummings, J. (1999). Expletives: Neurolinguistic
and neurobehavioral perspectives on swearing. Brain Research
Reviews, 31, 81–104.
Van Lancker, D., & Grafton, S. (1999a). PET activation studies
comparing speech tasks widely used in surgical mapping: Findings
in aphasic subjects. In Fifth international conference on functional
mapping of the human brain, Duesseldorf, Germany, June.
Van Lancker, D., & Grafton, S. (1999b). PET activation studies
comparing counting and naming in normal and aphasic subjects.
1999 Academy of Aphasia, Venice, Italy, October 23–26. Brain and
Language, 69, 434–437.
Votaw, J., Faber, T., Popp, C., Henry, T., Trudeau, J., Woodard, J.,
Mao, H., Hoffman, J., & Song, A. (1999). A confrontational
naming task produces congruent increases and decreases in PET
and fMRI. NeuroImage, 10, 347–356.
Wada, J., & Rasmussen, T. (1960). Intracarotid injection of sodium
amytal for the lateralization of cerebral speech dominance.
Experimental and clinical observations. Journal of Neurosurgery,
17, 266–282.
Wise, R., Hadar, U., Howard, D., & Patterson, K. (1991). Language
activation studies with positron emission tomography. CIBA
Foundation Symposium, 163, 218–228.
Woods, R. P., Grafton, S. T., Holmes, C. J., Cherry, S. R., &
Mazziotta, J. C. (1998a). Automated image registration: I. General
methods and intrasubject, intramodality validation. Journal of
Computer Assisted Tomography, 22, 139–152.
Woods, R. P., Grafton, S. T., Watson, J. D. G., Sicotte, N. L., &
Mazziotta, J. C. (1998b). Automated image registration: II.
Intersubject validation of linear and nonlinear models. Journal of
Computer Assisted Tomography, 22, 155–165.